I spent a really interesting day at the Operational Research Society’s Simulation Special Interest Group meeting at Surrey University today. The theme of the day was “Agent Based Simulation: are we there yet?” and we heard from a varied range of speakers from academia, the public sector and consultancy. My partner Dave Buxton (DSE Consulting) presented one of the papers.
Agent Based Simulation (ABS) has been employed for over two decades in theoretical work in the social sciences, but its use has been limited in comparison with other approaches to simulation such as Discrete Event Simulation and System Dynamics.
We heard presentations on ABS models that have already helped decision-makers in a range of application areas such as healthcare, economics and retail.
Some of the key points I took away included:
- ABS models are based on theoretical models, unlike many other OR simulations which are built on empirical models
- For models to have high credibility, they have to be built with the involvement of key stakeholders and provide specific, credible benefits
- ABS models can help with “learning” about the causes of a problem; they (generally) cannot be used as forecasting tools to give precise point predictions of a future state
- Building ABS models forces stakeholders to be precise about the theories they believe underpin the situation they are examining (e.g. it is no use saying “obesity will decrease”; you have to specify what will be happening to cause this outcome)
Finally, an interesting Catch-22 observation was made about all models/simulations: If the model tells stakeholders what they already know, then they will say they didn’t need the model; if it tells them something they didn’t know, they will say the model is wrong!
For more information on how Agent Based Simulation might help you solve strategic or tactical business problems, please contact me.
Download my article: “Simulation: a world of what if?”
I will post links to the presentations once they are available.